Release 1.0

• Overscheduling of possibly slow tasks.
• Fetching reduce intermediate data in parallel
• Better handling of large files for the S3 abstraction layer
• Better handling of large files for DFS
• Improved Dashboard layout
• Viewing of master and worker logs on the Dashboard
• Allow Map sizes to be configured

Release 0.6.0

• Add support for running Cerberus on AWS
• Implement DataAbstractionLayer for S3
• Dashboard improvements, allow for scheduling and canceling MapReduce jobs
• Worker performance and memory usage improvements
• Added state saving for distributed file system
• When running Cerberus using DFS task assignment will now prioritize workers that have some or all of the data for a task
• Increase reliability by re-doing map tasks if the worker with the intermediate data becomes unreachable to the master

Release 0.5.0

• Added Job and Task priority
• Created cluster dashboard
• Improved MapReduce algorithm by adding combiner functionality
• Added a distributed filesystem
• Added ability to cancel map reduce jobs

Release 0.4.0

Sprint 4 release

• Master
  • Finished master refactor
  • Changed input chunking to not write new files
  • Added data abstraction layer
  • Allow running multiple jobs simultaneously
  • Additional logging
  • General bug fixes
• Worker
  • Added support for new input chunking
  • Added data abstraction layer
  • Supported partitioning of reduce key space

Release 0.3.1

• Worker
  • Sharing of intermediate data between the workers.

Release 0.3

Sprint 3 release.

Release 0.2

Sprint 2 release

• Client Library

  • Added partitioning of reduce key space

• Master

  • Added state saving for recovery
  • Merged client GRPC Server with worker GRPC Server
  • Improved reliability
  • General bugfixes

• Worker

  • Changed the polling behavior to notifying the master when the task is complete
  • Supported partitioning of reduce key space

• CLI:

  • Display output directory
  • Make the output directory configurable

Release 0.1

Sprint 1 release.